Automated Machine Learning : Goals, Technicalities & Practical Insights
The field of automated machine learning (Auto ML) aims to develop methods that select & build suitable machine learning models without or with as little as possible human interventions.
In this presentation (accompanied with a demo), we shall review :
- The major goals of Auto ML in our machine learning pipelines and how it really helps Data Scientists to derive better models, parameters to improve both training & testing results.
- The major components in the pipeline including - Feature Engineering, Feature Selection, Model Selection, Hyper parameter optimization, Model evaluation, Performance evaluation & the iterative approach to continuously improve the process.
- The methods & mathematics available that make AutoML possible. And the merits/demerits of some of those methods as used in implementations.
- We shall also have a live demo of Auto-SkLearn and TPOT to get a good overview on how two of the most used frameworks are implemented.
- Evaluating frameworks for your implementation: We shall summarize the top questions to ask when selecting AutoML frameworks and how various frameworks compare with each other along with practical insights.
- Finally, we shall conclude on the value additions of AutoML in both business efficiency, turn around time, effective resource utilization and how it improves ML implementations for better bottom line/client implementations .
Outline/Structure of the Talk
* Introduction & Agenda (1m)
* Machine Learning Pipelines - Building & Deployment (2m)
*Goals of Auto ML (2m)
* Methods & Mathematics of Auto ML implementations (5m)
* Demo of TPOT/Auto-SkLearn(7m)
* Evaluating Auto ML Frameworks(2m)
* Conclusions (1m)
+ Understand Pipelines & Auto ML.
+ How Auto ML can help you?
+ How to choose the right AutoML framework for your needs
Data Scientists, ML Engineers, Leaders , Chief Data Officers, Chief Data Scientists
Prerequisites for Attendees
+ Understand foundations of Machine Learning Pipelines